What does TEO-CB-Auto-Env stand for?
Well, its short for
Teager Energy Operator Critical Band Autocorrelation Envelope.
That is the algorithm that
Advanced Lie Detector uses for analysis of speech. (Well, it is modified somewhat to work better with continuous speech). The graphic to the left is from a NATO research publication showing the tested accuracy of various methods of Voice stress detection and classification.
As shown in the graph, the TEO-CB-Auto-Env algorithm scores the highest with an accuracy rate of 94.2% with a 3.97 standard deviation in stress classification results.
Does this mean that with Advanced Lie Detector, I can detect lies with a 94.2% accuracy? No.
First of all, there is a difference between Stress and a Lie. One can be stressed and telling the truth, or one can lie and not be stressed. Usually, they go hand in hand, and that is the premise behind the Polygraph. In order to increase the correlation of stress to lies,
you should never expect Advanced Lie Detector to work on yourself. When you tell a lie, you are not stressed, because there are no consequences of lying to yourself. Ideally, Advanced Lie Detector should be used on someone else. With the more intimate and personal questions that tend to produce a higher stress when one is being deceitful.
Ok, so now that you understand the difference between Stress and a Lie.
Does that mean Advanced Lie Detector can detect Stressed Speech 94.2% of the time? Maybe.
The answer to that question is much more difficult. If you read through the
NATO research, one quickly realizes the results from the above graph are ideal LAB results. In fact, the researchers were not even feeding continuous speech into the recognition system, they manually cropped certain syllables and fed them into the system.
Stress detection on continuous speech is much more difficult for a variety of reasons.
